Learning to Collide: An Adaptive Safety-Critical Scenarios Generating Method

@article{Ding2020LearningTC,
  title={Learning to Collide: An Adaptive Safety-Critical Scenarios Generating Method},
  author={Wenhao Ding and Minjun Xu and Ding Zhao},
  journal={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  year={2020},
  pages={2243-2250}
}
Long-tail and rare event problems become crucial when autonomous driving algorithms are applied in the real world. For the purpose of evaluating systems in challenging settings, we propose a generative framework to create safety-critical scenarios for evaluating specific task algorithms. We first represent the traffic scenarios with a series of autoregressive building blocks and generate diverse scenarios by sampling from the joint distribution of these blocks. We then train the generative… Expand
Corner Case Generation and Analysis for Safety Assessment of Autonomous Vehicles
TLDR
A unified framework is proposed to generate corner cases for the decision-making systems based on Markov Decision Process, and the deep reinforcement learning techniques are applied to learn the behavior policy of BVs. Expand
Context-Aware Safe Reinforcement Learning for Non-Stationary Environments
TLDR
This paper uses a probabilistic latent variable model to achieve fast inference of the posterior environment transition distribution given the context data, and proposes the context-aware safe reinforcement learning (CASRL) method, a metal-earning framework to realize safe adaptation in non-stationary environments. Expand
Diverse Critical Interaction Generation for Planning and Planner Evaluation
TLDR
A styled generative model RouteGAN is proposed that generates diverse interactions by controlling the vehicles separately with desired styles by altering its style coefficients and can generate trajectories with different safety levels serve as an online planner. Expand
AdvSim: Generating Safety-Critical Scenarios for Self-Driving Vehicles
TLDR
This paper proposes AdvSim, an adversarial framework to generate safetycritical scenarios for any LiDAR-based autonomy system, and shows that the robustness and safety of these systems can be further improved by training them with scenarios generated by AdvSim. Expand
Neural Network Guided Evolutionary Fuzzing for Finding Traffic Violations of Autonomous Vehicles
TLDR
A new fuzz testing technique, called AutoFuzz, which can leverage widely-used av simulators’ API grammars to generate semantically and temporally valid complex driving scenarios (sequences of scenes) and reduces the traffic violations found in the new version of the av controller software. Expand
Multimodal Safety-Critical Scenarios Generation for Decision-Making Algorithms Evaluation
TLDR
A flow-based multimodal safety-critical scenario generator for evaluating decision-making algorithms and evaluates six Reinforcement Learning algorithms with their generated traffic scenarios and provides empirical conclusions about their robustness. Expand
A Modeled Approach for Online Adversarial Test of Operational Vehicle Safety
TLDR
A model-driven online feedback control policy for multiple POVs which propagates efficient adversarial trajectories while respecting traffic rules and other concerns formulated as an admissible state-action space is proposed. Expand
Behavior Varied Adversarial Pedestrian Generation for Autonomous Vehicle Testing
  • 2021
There exist several datasets for developing self-driving car methodolo1 gies. However, manually collected test datasets impose inherent limitations on 2 the variability of test cases, and it isExpand
Building Safer Autonomous Agents by Leveraging Risky Driving Behavior Knowledge
TLDR
This study generates multiple autonomous driving scenarios by creating new custom Markov Decision Process (MDP) environment iterations in the highway-env simulation package and trains model free learning agents with supplement information of risk prone driving scenarios and compares their performance with baseline agents. Expand
Detecting Safety Problems of Multi-Sensor Fusion in Autonomous Driving
TLDR
This work observed that the popular MSF methods in an industry-grade Advanced Driver-Assistance System (ADAS) can mislead the car control and result in serious safety hazards, and develops a novel evolutionary-based domain-specific search framework, FusionFuzz, for the efficient detection of fusion errors. Expand
...
1
2
...

References

SHOWING 1-10 OF 43 REFERENCES
CMTS: A Conditional Multiple Trajectory Synthesizer for Generating Safety-Critical Driving Scenarios
TLDR
It is concluded that the use of data generated by the Conditional Multiple Trajectory Synthesizer can improve the accuracy of trajectory predictions and autonomous vehicle safety. Expand
Adaptive Stress Testing for Autonomous Vehicles
TLDR
It is shown that DRL can find more likely failure scenarios than MCTS with fewer calls to the simulator and can be easily applied to other scenarios given the appropriate models of the vehicle and the environment. Expand
Scalable End-to-End Autonomous Vehicle Testing via Rare-event Simulation
TLDR
A simulation framework that can test an entire modern autonomous driving system, including, in particular, systems that employ deep-learning perception and control algorithms, is implemented. Expand
Efficient Autonomy Validation in Simulation with Adaptive Stress Testing
TLDR
By simulating an instance where an autonomous car drives while a pedestrian is crossing a road, it is demonstrated the solver is now capable of finding solutions for problems that would have previously been intractable. Expand
A New Multi-vehicle Trajectory Generator to Simulate Vehicle-to-Vehicle Encounters
TLDR
Comparison of the proposed MTG with $\beta$-VAE and InfoGAN demonstrates that the MTG has stronger capability to purposely generate rational vehicle-to-vehicle encounters through operating the disentangled latent codes. Expand
CARLA: An Open Urban Driving Simulator
TLDR
This work introduces CARLA, an open-source simulator for autonomous driving research, and uses it to study the performance of three approaches to autonomous driving: a classic modular pipeline, an end-to-end model trained via imitation learning, and an end to-end models trained via reinforcement learning. Expand
Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
TLDR
This paper proposes soft actor-critic, an off-policy actor-Critic deep RL algorithm based on the maximum entropy reinforcement learning framework, and achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off- policy methods. Expand
Accelerated Evaluation of Automated Vehicles Safety in Lane-Change Scenarios Based on Importance Sampling Techniques
  • Ding Zhao, H. Lam, +4 authors C. S. Pan
  • Computer Science, Medicine
  • IEEE Transactions on Intelligent Transportation Systems
  • 2017
TLDR
Frontal collision due to unsafe cut-ins is the target crash type of this paper and the cross-entropy method is used to recursively search for the optimal skewing parameters to accelerate the verification of AVs in simulations and controlled experiments. Expand
Accelerated Evaluation of Automated Vehicles in Car-Following Maneuvers
TLDR
The proposed techniques have great potential for accelerating the AV evaluation process, focusing on the car-following scenario, and show that test duration is reduced by a factor of 300 to 100 000 compared with the non-accelerated (naturalistic) evaluation. Expand
Adaptive stress testing of airborne collision avoidance systems
TLDR
This paper presents a scalable method to efficiently search for the most likely state trajectory leading to an event given only a simulator of a system using Monte Carlo Tree Search (MCTS), and presents results for both single and multi-threat encounters. Expand
...
1
2
3
4
5
...